Something for the Weekend - Part 261 - Know the Drill!

I decided I’d missed a trick when I did the DeWalt Atom car with Dave driving, and I had some credits I needed to burn because they reset tomorrow.  This time I used a 4 second loop of 720P in Firefly with a GIF in mind.  Despite using the same image for first and last frame I had to cut the last three frames in Photoshop. 

Then use Handbrake with the 480P preset that drops the size to 720x480 (16:9)

Check the Dimension tab

And most importantly the Video tab and drop the frame rate top 15fps (halve the file size)

Open the resulting file in Photoshop and use File > Export > Save for Web

I think the file size was around 7mb at that stage.  You can crop the frame and export again, or use Image Size to drop the size.  I wanted to keep the size so I reduced the colors to 64 which took the file size to 4.5Mb and below the forum’s 5mb limit.

It does lose some impact at that color depth, but if you double click the black and white swatches and set them to full black and full white, you win back some contrast.

The new challenge is up and running here : 

Come and take part 😊
Dave

My dyslexia, always words challenge round way wrong

Battle stations! All hands on deck!

One last upload (two really).  I wanted to create a looping background for the sporty car I made from the drill.  My idea was to use the same image for both first and last frame reference, and describe the moving background I was after.  It worked a treat, but there is a tiny glitch that is my error — I should have deleted the last frame when I stacked the video clips because there are two frames with no movement.  AFAIK Veo 3.1 is the only model that lets us upload first and last frames.  As before, I generated 720P and used Topaz Video to upres to 1080P 

Exporting even a single loop of the 720P produced a 60Mb GIF, and took some work to get below the forum’s new 5Mb limit.  I used Handbrake to reduce the video to 576P and 12fps, which exported to a 13Mb GIF You can downsize video inside Photoshop if you are making a GIF, and I had to reduce to 500px wide and 64 colors to get it down to 4Mb.  It was looking a bit flat but you can win back some contrast by double clicking the darkest tone in the color table and moving it nearer to full black.

The last GIF was made manually in Photoshop, but I was interested to see what Firefly could do. The prompt was simple and it only used 200 credits.  I used Premiere Pro again to reverse half thew clips to make the explosion effect bounce, but I also tried the Topaz Video slow-mow effect for the first time.  Topaz Video comes with Studio, but I have hardly used it and I wanted the practice.  The movement is a bit jerky I’m thinking. That might be me or might be the way it works.   Firefly produced the exploding drill on a white background.  To knock out the white, set the clips to Multiply in Effect Controls > Opacity.

They have dropped the maximum file size down from 10Mb to 5Mb and this a very unsatisfactory compromise in order to upload.  It’s reduced to 64 colors and 580 pixels wide, but I had to remove the nice fade that made the loop less abrupt.  It was all done in Photoshop apart from reversing the video, (it’s infinitely easier to animate a video timeline) 

May the drill be we you!

I’m struggling for inspiration this week. It’s probably down to the heatwave.

This is my younger brother Glyn. He is 74 now so a fair few years ago.  A while back I tried to get Firefly to make video of him riding the pig while it bucked like a rodeo bull, but it wouldn’t do it.  I have since realised that Firefly and Gemini 3.1 (which I use most of the time) will put up one of those orange ‘read the guidelines’ errors whenever it decides there are youngsters in the original image.  Using phrases like ‘young boy’ or ‘young girl’ in the prompt is going to cause the Ai to say no.

Nancy Sherwood recently put up a link to a FatherPhi video in which Phi Nguyen lies to Ais like ChatGTP.  The videos are hilarious and he so easily confuses the heck out of several Ai models.  Check him out. They are funny as heck.

So I wondered if we could cheat with the Photoshop partner Ai models.  In my first attempt I described Glyn as an ‘old man’ and got an error saying that the description did not match the content.  I tried again describing Glyn as a ‘young man’ which I thought was a believable compromise.  In my first attempt I used this old image of mine as a reference, and the Ai took it too far making it unrecognisable as Glyn and the farm. 

So I tried again without the reference using this prompt:

The young man standing next to the pig is to be riding the pig. The pig is bucking like a rodeo bull. The young man holds on to a rope tied around the pig with his right hand and waves his left hand in the air to help him balance. The background to be moved back to give the bucking pig more room but identifiable as the same location. The young man’s face to be unchanged. The pig to be the same as the original. The overall look to be exciting. The lighting to be dynamic and cinematic.

I am sure I could get it better than the above. Using Firefly to expand the background before using Gemini.  

I think we can bypass some of the Ai restraints by being careful with prompts.  You could add a moustache to the person in the original and remove it later.  I have other ideas that I am reluctant to share on an open forum, but they are basically about knowing that the Ai is not all powerful and can be tricked.  Check the FatherPhi channel to see what I mean.  

I am going to try again with Firefly to make a video of Glyn riding the pig. 

The Dentist.